Leveraging Machine Learning for Advanced Passive Sonar Tracking

The DON26BZ01-NV025 SBIR topic seeks to integrate cutting-edge machine learning into passive sonar processing to enhance tracking, classification, fusion, and localization in anti-submarine warfare systems. The project aims to develop advanced automation to detect, locate, classify, and correlate contacts across multiple sonar sensors and display surfaces, ultimately reducing operator workload. Specific performance improvements are targeted, including increased hold time ratio, reduced detection time, improved classification probability, decreased false alerts, enhanced track fusion, and a smaller area of uncertainty for localization. The effort is restricted under ITAR/EAR, requiring disclosure of foreign national involvement and strict adherence to U.S. export control laws. Phase I focuses on algorithm development, while Phase II involves implementation in a simulated environment using government-provided data, with work likely becoming classified. The contractor must be U.S.-owned and operated with secret level facility and personnel security clearances. Phase III will support transition to Navy use and explore dual-use applications in commercial surveillance and ISR uncrewed aerial vehicles.